I get the point about glass attenuation, but with regard to AR (anti-reflection?) coatings, are you referring to the surface through which light enters the prisms? If those surfaces are normal to the laser beam, would the reflected light not still be contained within the beam path, just as it is in the case of a Fabry-Perot resonator?

Yes, but at the wrong path length (and offset by some vector, if you had a TIR cube)